Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Priority
Acknowledgment is made of applicant’s claim for foreign priority under 35 U.S.C. 119 (a)-(d). The certified copy has been filed in parent Application No. CN2021-11583487.7, filed on December 22, 2021.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 07 Oct 2024 and 02 Apr 2025 are being considered by the examiner.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Regarding Claim 1
Step 1 – Whether the claim falls within any statutory category. See MPEP 2106.03
Claim 1 is a method claim and thus falls into one of the four categories of statutory subject matter.
Step 2A Prong 1 – Whether the claim recites a judicial exception. See MPEP 2106.04, subsection II
Regarding independent claim 1, the following limitations recite a judicial exception:
“performing, by a transmit end, first encoding processing on first data by using an encoder neural network, to obtain a first sent feature, wherein the first sent feature is related to a channel distribution dimension of an environment in which the transmit end is located”
[Mathematical Calculations / Concepts] – Encoding data using a neural network to extract and output a feature vector/tensor related to channel dimensions inherently relies on mathematical algorithms, matrix multiplications, and statistical vector transformations, which recite an abstract idea.
“performing, by the transmit end, second encoding processing on the first sent feature by using a matching layer, to obtain a first feature, wherein a dimension of the first feature is less than a dimension of the first sent feature”
[Mathematical Calculations / Concepts] – Performing second encoding processing using a matching layer to reduce feature dimensionality involves purely mathematical linear/nonlinear operations (such as matrix transformations), which recite an abstract idea.
Step 2A Prong 2 – Whether the claim recites additional elements that integrate the exception into a practical application of the exception? See MPEP 2106.04(d)
Regarding Claim 1, the claim recites the additional elements of:
“sending, by the transmit end, the first feature to a receive end.”
These ends are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
Sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than carry out the mathematical encoding calculations using generic communication components as a tool; thus, the claim as a whole does not integrate the abstract idea into a practical application.]
Step 2B – Whether the claim as a whole amounts to significantly more than the judicial exception? See MPEP 2106.05
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 2
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 2 is a dependent claim of 1, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim X, following limitations recite a judicial exception:
“updating, by the transmit end, a parameter of the matching layer based on a current channel of the transmit end”
[Mathematical Calculations] – calculating, adjusting, or optimizing numerical weight matrices or parameters within a neural network layer using numerical evaluations of channel conditions (e.g., error metrics or loss functions) represents pure mathematical algorithms and mathematical calculations, which recite an abstract idea.
“performing, by the transmit end, first encoding processing on second data by using the encoder neural network, to obtain a second sent feature”
[Mathematical Calculations] – executing neural network encoding calculations on second data vectors to generate numerical feature outputs relies on mathematical operations and matrix calculations, which recite an abstract idea.
“performing, by the transmit end, second encoding processing on the second sent feature by using the updated matching layer, to obtain a second feature”
[Mathematical Calculations] – executing matrix transformations on numerical features using updated weight parameters represents pure mathematical calculations and algorithmic evaluations, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 2, the claim recites additional elements of
“sending, by the transmit end, the second feature to the receive end, wherein the second feature is used by the receive end to obtain the second data”
The transmit end and the receive end are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
Sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 3
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 3 is a dependent claim of 2, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 3 does not have any abstract idea by itself, thus uses all the limitations of Claims 2
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 3, the claim recites additional elements of
“receiving, by the transmit end, first indication information from the receive end, wherein the first indication information indicates the transmit end to update the parameter of the matching layer”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
receiving data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 4
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 4 is a dependent claim of 2, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 4, following limitations recite a judicial exception:
“during the updating, by the transmit end, the parameter of the matching layer based on the current channel of the transmit end, a parameter of the encoder neural network is fixed”
[Mathematical Calculations] – holding or fixing numerical parameter values of an encoder neural network constant while calculating, adjusting, or optimizing weight matrices of a matching layer represents purely mathematical calculations, algorithmic optimization rules, and numerical matrix manipulations, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 4, the claim recites additional elements of
“the transmit end” and “neural network”
The transmit end and neural network are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 5
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 5 is a dependent claim of 2, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 5, following limitations recite a judicial exception:
“updating, by the transmit end, the parameter of the matching layer based on the current channel of the transmit end, a third sent feature, and a third received feature, wherein the third received feature comprises a feature obtained by the receive end by receiving, on a first channel, a third feature sent by the transmit end, the third feature comprises a feature obtained by the transmit end by performing second encoding processing on the third sent feature by using the matching layer, and the current channel of the transmit end is different from the first channel”
[Mathematical Calculations] – evaluating, calculating, or updating neural network weight matrices using stored historical numerical feature vectors (third sent feature vectors and third received feature vectors) and channel response parameters relies on pure mathematical operations, numerical algorithms, and matrix calculations, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 5, the claim recites additional elements of
“transmit end” and “receive end”
The transmit end and the receive end are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
The claim does not provide an inventive concept (significantly more than the abstract idea). The claim is ineligible.
As explained above, the additional element [1] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)). This limitation remains a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 6
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 6 is a dependent claim of 5, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 6, following limitations recite a judicial exception:
“determining, by the transmit end, the current channel of the transmit end based on the first information”
[Mathematical Calculations] – evaluating, calculating, or updating neural network weight matrices using stored historical numerical feature vectors (third sent feature vectors and third received feature vectors) and channel response parameters relies on pure mathematical operations, numerical algorithms, and matrix calculations, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 6, the claim recites additional elements of
“obtaining, by the transmit end”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
obtaining data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g))
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 7-12
Claims 7-12 have similar limitations (Transmit end implementing methods of Claim 1-6) of Claims 1-6, respectively. For the reasons described above with respect to Claims 1-6, these judicial exceptions are not meaningfully integrated into a practical application, or significantly more than the abstract ideas. The claims do not provide anything more than the abstract ideas of mental processes and mathematical calculations that are practically capable of being performed with the assistance of pen and paper. Therefore, claims 7-12 also recite abstract ideas that do not integrate into a practical application or amount to significantly more than judicial exception, and thus are rejected under U.S.C. 101.
Regarding Claim 13
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 13 is a dependent claim of 12, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 13, following limitations recite a judicial exception:
“wherein the first information comprises channel information or received feature offset information from the receive end; and the channel information represents related information of the current channel of the transmit end, the received feature offset information represents a difference between the third received feature and a fourth received feature, the third received feature comprises the feature obtained by the receive end by receiving, on the first channel, the third feature sent by the transmit end, and the fourth received feature comprises a feature obtained by the receive end by receiving, on the current channel, the third feature sent by the transmit end”
[Mathematical Calculations / Mental Process] – calculating vector differences between baseline received features and current channel received features, or extracting channel state metrics represents mathematical formulas, algorithmic evaluations, and numerical data manipulation, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 13, the claim recites additional elements of
“sent by the transmit end” and “Obtained by receive end”
The transmit end and the receive end are recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
sending and obtaining data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 14
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 14 is an apparatus claim thus it falls into one of the four categories of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding independent claim 14, following limitations recite a judicial exception:
“performing decoding processing on the first received feature by using a decoder neural network, to obtain the first data, wherein the decoder neural network and the matching layer are obtained through independent training”
[Mathematical Calculations] – executing neural network decoding transformations via matrix multiplications to reconstruct original data vectors from numerical feature vectors relies on mathematical concepts, algorithms, and numerical calculations, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 14, the claim recites additional elements of
“receiving a first received feature from a transmit end, wherein the first received feature comprises a feature that is received by a receive end and that is obtained by transmitting, through a channel, a first feature sent by the transmit end, the first feature is obtained by the transmit end by performing encoding processing on a first sent feature by using a matching layer, and the first sent feature is obtained by an encoder neural network of the transmit end by performing encoding processing on first data”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
receiving data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 15
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 15 is a dependent claim of 14, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 15 does not have any abstract idea by itself, thus uses all the limitations of Claim 14.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 15, the claim recites additional elements of
“sending first indication information to the transmit end, wherein the first indication information indicates the transmit end to update a parameter of the matching layer”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 16
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 16 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 16, following limitations recite a judicial exception:
“a parameter representing a channel change degree is less than or equal to a first threshold”
[Mathematical Calculations] – evaluating and comparing a measured numerical channel variation parameter (x) against a decision threshold value to determine conditional execution represents mathematical comparisons, algorithmic evaluation rules, and logical data processing, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 16, the claim recites additional elements of
“sending first indication information to the transmit end”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 17
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 17 is a dependent claim of 14, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
Regarding dependent claim 17, following limitations recite a judicial exception:
“a parameter representing a channel change degree is greater than a first threshold, wherein the second indication information indicates the transmit end to retrain the encoder neural network”
[Mathematical Calculations] – numerically evaluating channel degradation parameters against a boundary condition and transmitting an algorithmic trigger signal represents mathematical calculations, algorithmic rules, and generic data transmission, which recite an abstract idea.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 17, the claim recites additional elements of
“sending second indication information to the transmit end”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 18
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 18 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 18 does not have any abstract idea by itself, thus uses all the limitations of Claim 15.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 18, the claim recites additional elements of
“receiving third indication information to the transmit end”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
receiving data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 19
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 19 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 19 does not have any abstract idea by itself, thus uses all the limitations of Claim 15.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 19, the claim recites additional elements of
“receiving fourth indication information from the transmit end, wherein the fourth indication information indicates that training of the encoder neural network is completed”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
receiving data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Regarding Claim 20
Step 1 – whether the claim falls within any statutory category. See MPEP 2016.03
Claim 20 is a dependent claim of 15, thus it falls within the same category of statutory subject matter.
Step 2A Prong 1 – whether the claim recites a judicial exception. See MPEP 2106.04, subsection II.
As Claim 20 does not have any abstract idea by itself, thus uses all the limitations of Claim 15.
Step 2A Prong 2 – whether the claim recites additional elements that integrate the exception into a practical application of the exception?
Regarding Claim 20, the claim recites additional elements of
“sending first information to the transmit end, wherein the first information comprises information for updating the parameter of the matching layer”
The transmit end is recited at a high level of generality and is merely adding words “apply it” to the judicial exception. (see MPEP 2106.05(f))
sending data over the network is merely data gathering recited at a high level of generality, thus is insignificant extra-solution activity (See MPEP 2106.05(g)).
[Even when viewed in combination, the additional elements do no more than automate the mental processes that a person could perform, using computer components as a tool, thus the claim as a whole does not integrate into a practical application.]
Step 2B – whether the claim as a whole amount to significantly more than the judicial exception? I.e. Are there any additional elements (features/limitations/step) recited in the claim beyond the abstract idea?
As explained above, the additional element [1a] is considered a mere instruction to apply an exception to the generic computer components or machine-learning components that simply run mathematical calculations and mental processes (see MPEP 2106.05(f)).
The additional element [1b] is considered an insignificant extra solution activity and at best the equivalent of a mere data gathering recited at a high level of generality and amount to receiving or transmitting data over a network, which is well-understood, routine, conventional activity (See MPEP 2106.05(d), subsection II).
These limitations remain insignificant extra-solution activity and a mere instruction to apply an exception even upon reconsideration. Even when considered in combination, the additional element represents insignificant extra-solution activity and a mere instruction to apply an exception, which cannot provide an inventive concept.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
Claims 1, 2, 4, 7, 8, 10 are rejected under 35 U.S.C. 103 as being unpatentable over Xie et al. (Xie), Non-Patent Literature, “Deep Learning Enabled Semantic Communication Systems”, published in 19 Apr 2021, Pages: 13, in view of Park et al. (Park), Non-Patent Literature (listed in IDS filed on 07 Oct 2024), “Meta-Learning to Communicate: Fast End-To-End Training for Fading Channels”, published on 2020, Pages: 5.
As to independent Claim 1,
Xie teaches a communication method, comprising:
performing, by a transmit end, first encoding processing on first data by using an encoder neural network, to obtain a first sent feature (
Xie, Pg3, Right Column(RC hereinafter), Subsection A, Paragraph2(P2), Lines1-5, “Particularly, we assume that the input of the DeepSC is a sentence, s=[w_1, w_2, …, w_L], where w_l represents the l-th word in the sentence… The transmitter consists of two parts, named semantic encoder and channel encoder, to extract the semantic information from s..."
Pg6, RC, P1, Lines6-11,
PNG
media_image1.png
137
487
media_image1.png
Greyscale
wherein Xie explicitly discloses a transmitter (transmit end) that receives input text sentences s (first data) and processes them through an embedding layer and a Transformer-based semantic encoder (encoder neural network) during an initial encoding stage (first encoding processing) to produce an extracted semantic feature tensor
PNG
media_image2.png
22
95
media_image2.png
Greyscale
(first sent feature));
performing, by the transmit end, second encoding processing on the first sent feature by using a matching layer, to obtain a first feature, wherein a dimension of the first feature is less than a dimension of the first sent feature (
Pg5, RC, SubsectionA, Lines1-5, “Particularly, the transmitter consists of a semantic encoder to extract the semantic features from the texts to be transmitted and a channel encoder to generate symbols to facilitate the transmission subsequently"
Pg6, RC, P1, Lines8-13, "Then, pass the semantic encoder layer to obtain
PNG
media_image2.png
22
95
media_image2.png
Greyscale
, the semantic information conveyed by S, where V is the dimension of Transformer encoder's output. Then, M is encoded into symbols X to cope with the effects from the physical channel, where
PNG
media_image3.png
27
123
media_image3.png
Greyscale
"
Pg8, Tabl1, "Transmitter (Encoder): 3x Transformer Encoder (Units: 128 (8 heads)) ... Dense (Units: 256) ... Dense (Units: 16)"
PNG
media_image4.png
243
480
media_image4.png
Greyscale
wherein Xie explicitly discloses a transmitter (transmit end) that receives the semantic feature tensor M (first sent feature) output from the Transformer semantic encoder and processes it through dense layers of a channel encoder (Under the Broadest Reasonable Interpretation (BRI), a matching layer is reasonably interpreted in light of the specification to encompass any neural network layer or set of dense layers that further processes/encodes features for channel transmission), during a subsequent encoding stage (second encoding processing) to obtain transmitted symbol tensor X (first feature). Furthermore, Xie explicitly discloses that the feature dimension is reduced during this second encoding stage, as the input semantic feature M has a Transformer output dimension of V=128, whereas the channel encoder dense layers output a representation constrained to 16 units (or 2N per word). Thus, Xie directly teaches that the dimension of the first feature (X, 16 units) is less than the dimension of the first sent feature (M, 128 units))
sending, by the transmit end, the first feature to a receive end (
Xie, Pg7, Algorithm2 & 3, Line6, “Transmit X over the channel”
Pg6, RC, P2, Lines5-6, “At the receiver, distorted symbols Y are received and then decoded by the channel decoder layer…”
wherein Xie explicitly discloses transmitting the generated symbol representation X (first feature) from the transmitter over the physical channel to the receiver (receive end).)
However, Xie does not explicitly teach:
wherein the first sent feature is related to a channel distribution dimension of an environment in which the transmit end is located
In the same field of endeavor, Park teaches this feature (
Park, Pg5076, Section3.1, Lines2-5, “Under joint training, we wish to find a unique solution parameter
PNG
media_image5.png
24
14
media_image5.png
Greyscale
that works well on all channels in a set
PNG
media_image6.png
31
136
media_image6.png
Greyscale
sampled from a given fading distribution"
Pg5076, Section2, P2, Lines4-8,
PNG
media_image7.png
135
562
media_image7.png
Greyscale
wherein Park teaches training an encoder neural network (
PNG
media_image8.png
26
30
media_image8.png
Greyscale
) over a comprehensive set of channel state samples
PNG
media_image6.png
31
136
media_image6.png
Greyscale
derived from a fading channel distribution, thereby directly configuring the encoder's generated output feature (x) to be dependent upon and related to the channel distribution dimension of the environment, as the network is trained over a fading distribution set H capturing the environmental channel characteristics.)
Xie and Park are analogous to the claimed invention as they are from the same fields of endeavor of deep learning-based physical layer wireless communication systems and end-to-end autoencoder transceivers. Therefore, it would have been obvious to a person having ordinary skill in the art, before the effective filing date, to combine the Transformer-based semantic autoencoder transmitter architecture of Xie with the channel-distribution-aware joint training framework of Park.
The motivation is taught by Park (
Park, Pg5076, Section3.1, Lines2-5, “Under joint training, we wish to find a unique solution parameter
PNG
media_image5.png
24
14
media_image5.png
Greyscale
that works well on all channels in a set
PNG
media_image6.png
31
136
media_image6.png
Greyscale
sampled from a given fading distribution"
) such that training Xie’s semantic encoder across a sampled fading channel distribution set H conditions the extracted semantic feature (first sent feature M) on the underlying environmental channel characteristics, thereby ensuring robust semantic decoding performance and link reliability across varying fading channels without requiring structural alterations to Xie’s transmitter architecture.
As to dependent Claim 2,
The combination of Xie and Park teaches all the limitations of Claim 1 as set forth above.
Xie further teaches
updating, by the transmit end, a parameter of the matching layer based on a current channel of the transmit end (
Xie, Pg7, RC, Lines5-7, “For different communication environments, we redesign and train part of the channel encoder and decoder layers and freeze the semantic encoder and decoder layers.”
Pg7, Figure 5,
PNG
media_image9.png
297
481
media_image9.png
Greyscale
Pg7, Algorithm4, Second Function,
PNG
media_image10.png
115
484
media_image10.png
Greyscale
Wherein Xie explicitly discloses that when the communication environment or channel condition changes (i.e., based on different channel parameters / current channel), the transmit end updates (re-trains) the parameter set
PNG
media_image11.png
10
13
media_image11.png
Greyscale
of the channel encoder
PNG
media_image12.png
26
49
media_image12.png
Greyscale
(which corresponds to the claimed matching layer under Broadest Reasonable Interpretation) while keeping the semantic encoder
PNG
media_image13.png
26
47
media_image13.png
Greyscale
frozen)
performing, by the transmit end, first encoding processing on second data by using the encoder neural network, to obtain a second sent feature (
Xie, Pg6, RC, P1, Lines6-11,
PNG
media_image1.png
137
487
media_image1.png
Greyscale
Pg7, Algorithm2 & 3, Lines3-4,
PNG
media_image14.png
66
255
media_image14.png
Greyscale
wherein Xie explicitly discloses that the transmit end receives subsequent text sentences S (second data) and processes them through the frozen Transformer-based semantic encoder
PNG
media_image15.png
25
29
media_image15.png
Greyscale
(encoder neural network) during an initial encoding stage (first encoding processing) to extract a corresponding semantic feature tensor M (second sent feature))
performing, by the transmit end, second encoding processing on the second sent feature by using the updated matching layer, to obtain a second feature (
Xie, Pg6, RC, P1, Lines11-13, “Then, M is encoded into symbols X to cope with the effects from the physical channel, where
PNG
media_image3.png
27
123
media_image3.png
Greyscale
"
Pg7, Algorithm2 & 3, Line5,
PNG
media_image16.png
17
174
media_image16.png
Greyscale
wherein Xie explicitly discloses that the transmit end processes the extracted semantic feature tensor M (second sent feature) through the re-trained/updated channel encoder
PNG
media_image17.png
22
28
media_image17.png
Greyscale
(updated matching layer) during a subsequent encoding stage (second encoding processing) to obtain the updated transmitted symbol representation X (second feature)); and
sending, by the transmit end, the second feature to the receive end (
Xie, Pg7, Algorithm2 & 3, Line6, “Transmit X over the channel”
Pg6, RC, P2, Lines5-6, “At the receiver, distorted symbols Y are received and then decoded by the channel decoder layer…”
wherein Xie explicitly discloses transmitting the generated symbol representation X (second feature) from the transmit end over the physical channel to the receive end.)
As to dependent Claim 4,
The combination of Xie and Park teaches all the limitations of Claim 2 as set forth above.
Xie further teaches the method of claim2,
wherein a parameter of the encoder neural network remains unchanged in a process in which the transmit end updates the parameter of the matching layer (
Xie, Pg7, Algorithm4, Second Function, Lines9-11
PNG
media_image10.png
115
484
media_image10.png
Greyscale
wherein Xie explicitly discloses that when adapting to different channel conditions by re-training / updating the parameter set
PNG
media_image11.png
10
13
media_image11.png
Greyscale
of the channel encoder
PNG
media_image12.png
26
49
media_image12.png
Greyscale
(which corresponds to the claimed matching layer under Broadest Reasonable Interpretation), the parameters of the semantic encoder
PNG
media_image13.png
26
47
media_image13.png
Greyscale
(which corresponds to the claimed encoder neural network) are frozen (remain unchanged). Thus, Xie directly teaches that the parameter of the encoder neural network remains unchanged in the process in which the transmit end updates the parameter of the matching layer.)
As to independent Claim 7,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 1 and thus rejected under the same rationale.
As to dependent Claim 8,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 2 and thus rejected under the same rationale.
As to dependent Claim 10,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 4 and thus rejected under the same rationale.
Claims 3, 5, 6, 9, 11, 12, 13 are rejected under 35 U.S.C. 103 as being unpatentable over Xie and Park, as mentioned in Claim 2, in further view of Dorner et al. (Dorner), Non-Patent Literature, “WGAN-based Autoencoder Training Over-the-air”, published in 5 March 2020, Pages: 5.
As to dependent Claim 3,
The combination of Xie and Park teaches all the limitations of Claim 2 as set forth above.
However, both Xie and Park do not explicitly teach
receiving, by the transmit end, first indication information from the receive end, wherein the first indication information indicates the transmit end to update the parameter of the matching layer
In the same field of endeavor, Dorner teaches this feature (
Dorner, Pg1, RC, P2, Lines 9-11, “...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter...”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses a wireless communication autoencoder system where channel training/evaluation is performed at the receive end (or local receiver instance) and feedback information comprising updated transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
is sent back from the receive end to the transmit end to instruct/indicate the transmit end to update its neural network parameters (matching layer / transmitter parameters) to adapt to the physical channel environment.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the claimed “first indication information” is reasonably interpreted to encompass any feedback signal, control message, or updated parameter data
PNG
media_image18.png
29
30
media_image18.png
Greyscale
sent from a receive end that informs, instructs, or causes the transmit end to update or refresh its neural network parameters. Furthermore, applying Dorner’s receiver-to-transmitter weight update feedback mechanism to the split-autoencoder architecture of Xie (which maintains a frozen semantic encoder
PNG
media_image15.png
25
29
media_image15.png
Greyscale
and an adaptable channel encoder
PNG
media_image17.png
22
28
media_image17.png
Greyscale
) directly results in indicating the transmit end to specifically update the parameters of its channel encoder (corresponding to the claimed matching layer).)
Xie, Park, and Dorner are analogous to the claimed invention as they are from the same field of endeavor of deep learning-based physical layer wireless communication systems and end-to-end autoencoders. Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to combine the Xie-Park framework with the receiver-to-transmitter weight update feedback and indication mechanism of Dorner.
The motivation is as recited by Dorner (
Dorner, Pg1, RC, P2, Lines6-12, “Therefore, we show that training of GANs only requires a single dataset that can be collected in a one-shot transmission while this is not possible for RL-based training. Thus, training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter, while in RL training, a continuous feedback of the reward must be ensured”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel”
) such that transmitting updated weights/indication information from the receive end back to the transmit end enables the receive end to dictate, trigger, or supply transmitter parameter updates without requiring continuous real-time closed-loop reward feedback, thereby dramatically reducing feedback signaling overhead, transmission latency, and operational complexity while maintaining optimal end-to-end communication performance across dynamic channels.
As to dependent Claim 5,
The combination of Xie and Park teaches all the limitations of Claim 2 as set forth above.
However, both Xie and Park do not explicitly teach
updating, by the transmit end, the parameter of the matching layer based on the current channel of the transmit end, a third sent feature, and a third received feature, wherein the third received feature comprises a feature obtained by the receive end by receiving, on a first channel, a third feature sent by the transmit end, the third feature comprises a feature obtained by the transmit end by performing second encoding processing on the third sent feature by using the matching layer, and the current channel of the transmit end is different from the first channel.
In the same field of endeavor, Dorner teaches this feature (
Dorner, Pg1, RC, P2, Lines7-10, “...training of GANs only requires a single dataset that can be collected in a one-shot transmission while this is not possible for RL-based training. Thus, training can be fully done at the receiver...”
Pg2, RC, Paragraph 2, Lines 5-11, “...the transmitter generates a batch of symbols x, which are then sent over the actual channel and the received symbols y are recorded at the receiver. The receiver weights
PNG
media_image19.png
21
26
media_image19.png
Greyscale
can then easily be trained in a supervised fashion using SGD on loss...”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses collecting and recording a dataset comprising previously sent symbol features (x, corresponding to the claimed third sent feature) and corresponding received signal features (y_r, corresponding to the claimed third received feature) transmitted over an initial measurement channel (first channel), and utilizing this recorded dataset of sent and received features to train/update the transmitter neural network weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
, corresponding to the matching layer parameters under BRI) to adapt the transmitter parameters for deployment on a dynamic or different current channel condition.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the claimed “third sent feature” and “third received feature” encompass any recorded transmit symbol vectors and corresponding channel-affected received symbol/IQ sample vectors stored as a training dataset. Furthermore, updating the transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
based on the recorded dataset collected over a first channel measurement to configure the transmitter for current channel operation directly corresponds to updating the matching layer parameter based on a current channel, a third sent feature, and a third received feature. To the extent, applying the updated weights to refresh local parameters at the transmit end constitutes "updating by the transmit end" under BRI; alternatively, relocating or executing the dataset-based parameter optimization algorithm directly at the transmit end using fed-back feature samples represents a conventional design choice well within the routine capabilities of a POSITA.)
Xie, Park, and Dorner are analogous to the claimed invention as they are from the same field of endeavor of deep learning-based physical layer wireless communication systems and end-to-end autoencoders. Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to combine the Xie-Park framework with the recorded sent/received feature dataset-based parameter update mechanism of Dorner.
The motivation is as recited by Dorner (
Dorner, Pg1, RC, P2, Lines6-12, “Therefore, we show that training of GANs only requires a single dataset that can be collected in a one-shot transmission while this is not possible for RL-based training. Thus, training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter, while in RL training, a continuous feedback of the reward must be ensured”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel”
) such that collecting a single dataset of sent and received features over a first channel and updating the transmitter parameters offline eliminates the need for continuous closed-loop feedback and real-time retraining during dynamic operation, thereby dramatically reducing feedback signaling overhead, transmission latency, and operational complexity while maintaining optimal end-to-end communication performance across dynamic, time-varying wireless channels.
As to dependent Claim 6,
The combination of Xie, Park and Dorner teaches all the limitations of Claim 5 as set forth above.
Dorner further teaches the method according to claim 5, wherein before the updating, by the transmit end, the parameter of the matching layer based on the current channel of the transmit end, the method further comprises:
obtaining, by the transmit end, first information; and determining, by the transmit end, the current channel of the transmit end based on the first information (
Dorner, Pg3, LC, Lines16-22, “To estimate the SNR(signal-to-noise-ratio) required by the demapper, we first calculate the error vector mangnitude (EVM) between the originally sent symbols and the equalized received symbols. We then calculate an average SNR per sub-carrier, defined by the mean over the EVM, and feed this SNR estimation to the demapper of the corresponding sub-carrier.”
Pg4, RC, Section IV, Lines10-13, “...we needed to add a pre-processing step that zero-forces the phase
PNG
media_image20.png
19
20
media_image20.png
Greyscale
of the first sub-carrier in time domain, as due to slight carrier frequency offset (CFO) we noticed random
PNG
media_image20.png
19
20
media_image20.png
Greyscale
for different measurements.”
Pg4, Fig. 4, “Measured OTA and generated frequency response in magnitude and phase.”
PNG
media_image21.png
231
452
media_image21.png
Greyscale
wherein Dorner explicitly discloses that prior to updating/finetuning neural network parameters (or matching layer parameters), the system obtains feedback/measurement data (first information, e.g., EVM-based SNR estimates, CFO phase offset parameters
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, or measured OTA IQ samples) and determines/estimates the current channel state (current channel, e.g., subcarrier frequency response and SNR per subcarrier) based on this obtained information.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the claimed “first information” is reasonably interpreted to encompass any channel measurement parameter, EVM/SNR metrics, phase offset data, or recorded IQ samples. Furthermore, obtaining per-subcarrier SNR estimates or phase offsets
PNG
media_image20.png
19
20
media_image20.png
Greyscale
to determine current channel response prior to updating network parameters directly corresponds to obtaining first information and determining the current channel based on the first information.)
Xie, Park, and Dorner are analogous to the claimed invention as they are from the same field of endeavor of deep learning-based physical layer wireless communication systems and end-to-end autoencoders. Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to combine the Xie-Park-Dorner framework with the channel information obtaining and channel determining mechanism of Dorner.
The motivation is as recited by Dorner (
Dorner, Pg1, RC, P2, Lines7-10, “...training of GANs only requires a single dataset that can be collected in a one-shot transmission while this is not possible for RL-based training. Thus, training can be fully done at the receiver...”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
Pg5, Section V, Lines4-6, “...GANs benefits from a simpler data acquisition as the whole training set can be collected one-shot for constant transmitter weights”
) such that obtaining first information (such as EVM/SNR measurements or CFO phase offsets) and determining the current channel condition prior to executing matching layer parameter updates enables the system to collect training data in a simplified one-shot transmission and execute parameter updates offline, thereby dramatically reducing feedback signaling overhead, transmission latency, and deployment operational complexity of end-to-end training over the actual channel while maintaining optimal communication performance across dynamic, time-varying wireless channels.
As to dependent Claim 9,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 3 and thus rejected under the same rationale.
As to dependent Claim 11,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 5 and thus rejected under the same rationale.
As to dependent Claim 12,
it is an apparatus (transmitter) claim that contains similar limitations of Claim 6 and thus rejected under the same rationale.
As to dependent Claim 13,
The combination of Xie, Park, and Dorner teaches all the limitations of Claim 12 as set forth above.
However, both Xie and Park do not explicitly teach
the apparatus according to claim 12, wherein the first information comprises channel information or received feature offset information from the receive end; and the channel information represents related information of the current channel of the transmit end, the received feature offset information represents a difference between the third received feature and a fourth received feature, the third received feature comprises the feature obtained by the receive end by receiving, on the first channel, the third feature sent by the transmit end, and the fourth received feature comprises a feature obtained by the receive end by receiving, on the current channel, the third feature sent by the transmit end
In the same field of endeavor, Dorner further teaches (
Dorner, Pg3, Left Column, Section II-B, Lines 18-25, “To estimate the SNR required by the demapper, we first calculate the error vector magnitude (EVM) between the originally sent symbols and the equalized received symbols. We then calculate an average SNR per sub-carrier, defined by the mean over the EVM, and feed this SNR estimation to the demapper of the corresponding sub-carrier.”
Pg4, RC, Section IV, Lines10-13, “...we needed to add a pre-processing step that zero-forces the phase
PNG
media_image20.png
19
20
media_image20.png
Greyscale
of the first sub-carrier in time domain, as due to slight carrier frequency offset (CFO) we noticed random
PNG
media_image20.png
19
20
media_image20.png
Greyscale
for different measurements.”
Pg4, Fig. 4, “Measured OTA and generated frequency response in magnitude and phase.”
Pg2, Left Column, Section II-A, Paragraph 2, Lines 5-11, “...the transmitter generates a batch of symbols x, which are then sent over the actual channel and the received symbols y are recorded at the receiver.”
wherein Dorner explicitly discloses obtaining feedback data (first information) from the receive end, which comprises either: (1) Channel information representing related information of the current channel of the transmit end (e.g., per-subcarrier EVM-based SNR estimates, CFO phase offset parameters
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, or measured OTA channel frequency responses in magnitude and phase); OR (2) Received feature offset information representing a difference between a third received feature and a fourth received feature, wherein a baseline recorded received symbol/IQ sample vector y_r obtained on an initial measurement channel (first channel) for a transmitted feature x (third feature) corresponds to the claimed third received feature, a current received symbol vector y' obtained on a dynamic current channel for the same transmitted feature x corresponds to the claimed fourth received feature, and the difference metric computed therebetween (e.g., Error Vector Magnitude (EVM) or symbol error vectors) represents the difference between the third received feature and the fourth received feature.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification (specifically paragraph [0134] reciting
PNG
media_image22.png
30
112
media_image22.png
Greyscale
), the claimed “received feature offset information” encompasses any error vector magnitude (EVM), error metrics, or feature difference values calculated between baseline/reference channel receptions and current channel receptions. Furthermore, providing either per-subcarrier channel measurements (CFO phase offsets/frequency response) or EVM-based feature offset metrics from a receive end directly satisfies the alternative limitations of Claim 13.
Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to configure the apparatus of Claim 12 such that the first information fed back from the receive end comprises channel information (e.g., CFO phase offsets or subcarrier frequency responses) or received feature offset information (e.g., EVM or symbol error vector differences between baseline and current channel receptions) as taught by Dorner, in order to enable precise current channel identification and efficient offline parameter adaptation at the transmit end.)
Claims 14-20 are rejected under 35 U.S.C. 103 as being unpatentable over Xie et al. (Xie), Non-Patent Literature, “Deep Learning Enabled Semantic Communication Systems”, published in 19 Apr 2021, Pages: 13, in view of Dorner et al. (Dorner), Non-Patent Literature, “WGAN-based Autoencoder Training Over-the-air”, published in 5 March 2020, Pages: 5.
As to independent Claim 14,
Xie teaches a communication apparatus, comprising at least one processor, and one or more memories coupled to the at least one processor and storing programming instructions for execution by the at least one processor to perform operations comprising (
Xie, Pg6, RC, Section III, Lines1-9, "The considered system model consists of two levels: semantic level and transmission level... Overall, we consider an intelligent E2E communication system with the stochastic physical channel, where the transmitter and the receiver have certain background knowledge..."
Pg 8, Table I, "Receiver (Decoder): Dense (Units: 256) ... Dense (Units: 128) ... 3x Transformer Decoder (Units: 128 (8 heads))"
PNG
media_image4.png
243
480
media_image4.png
Greyscale
wherein Xie explicitly discloses a communication receiver apparatus equipped with hardware processors and memory storing neural network parameters and execution instructions to implement deep learning decoding operations);
receiving a first received feature from a transmit end, wherein the first received feature comprises a feature that is received by a receive end and that is obtained by transmitting, through a channel, a first feature sent by the transmit end (
Xie, Pg6, RC, Paragraph2, Lines5-6, "At the receiver, distorted symbols Y are received and then decoded by the channel decoder layer..."
Pg 7, Algorithm 2 & 3, Line 6, "Transmit X over the channel" & Line 8, "Receive Y"
wherein the distorted symbol vector Y corresponds to the claimed first received feature, the transmitted symbol vector X corresponds to the claimed first feature, and transmitting over the physical channel to receive Y at the receiver corresponds to receiving the first received feature obtained by transmitting the first feature through a channel);
first feature is obtained by the transmit end by performing encoding processing on a first sent feature by using a matching layer (
Xie, Pg6, RC, P1, Lines6-11,
PNG
media_image1.png
137
487
media_image1.png
Greyscale
Xie, Pg6, RC, P1, Lines11-13, “Then, M is encoded into symbols X to cope with the effects from the physical channel, where
PNG
media_image3.png
27
123
media_image3.png
Greyscale
"
wherein dense channel encoder layers encoding the semantic feature tensor M (first sent feature) into transmitted symbol vector X (first feature) correspond to performing encoding processing on a first sent feature using a matching layer under Broadest Reasonable Interpretation (BRI));
the first sent feature is obtained by an encoder neural network of the transmit end by performing encoding processing on first data (
Xie, Pg3, Right Column, Subsection A, Paragraph2(P2), Lines1-5, “Particularly, we assume that the input of the DeepSC is a sentence, s=[w_1, w_2, …, w_L], where w_l represents the l-th word in the sentence… The transmitter consists of two parts, named semantic encoder and channel encoder, to extract the semantic information from s..."
Xie, Pg6, RC, P1, Lines6-11,
PNG
media_image1.png
137
487
media_image1.png
Greyscale
wherein the input text sentence s corresponds to the claimed first data, the Transformer semantic encoder corresponds to the claimed encoder neural network, and the extracted semantic feature tensor M corresponds to the claimed first sent feature);
performing decoding processing on the first received feature by using a decoder neural network, to obtain the first data (
Xie, Pg 4, LC, Equation (3),
PNG
media_image23.png
198
483
media_image23.png
Greyscale
Pg 6, RC, P2, Lines5-8, "At the receiver, distorted symbols Y are received and then decoded by the channel decoder layer... Afterwards, the transmitted sentences are estimated by the semantic decoder layer"
wherein decoding received distorted symbols Y using channel decoder
PNG
media_image24.png
32
45
media_image24.png
Greyscale
and semantic decoder
PNG
media_image25.png
35
41
media_image25.png
Greyscale
to obtain recovered sentence
PNG
media_image26.png
21
15
media_image26.png
Greyscale
directly corresponds to performing decoding processing on the first received feature using a decoder neural network to obtain the first data.)
However, Xie does not explicitly teach:
wherein the decoder neural network and the matching layer are obtained through independent training.
In the same field of endeavor, Dorner teaches this feature (
Dorner, Pg1, RC, P2, Lines9-11, "...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter..."
Pg2, LC, Section II, P2, Lines10-18, "In the case of a channel model with known channel gradient, we simply train in an unsupervised end-to-end learning fashion... However, after deployment on an actual channel, the channel gradient is unknown and only
PNG
media_image27.png
21
24
media_image27.png
Greyscale
can be updated straightforwardly using backpropagation, described as finetuning... The receiver weights
PNG
media_image27.png
21
24
media_image27.png
Greyscale
can then easily be trained in a supervised fashion using SGD on loss (2)..."
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses a decoupled/independent training mechanism for physical layer autoencoders, wherein training or updating the receiver decoder neural network weights (
PNG
media_image27.png
21
24
media_image27.png
Greyscale
) and training or updating the transmitter neural network/matching layer weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) are executed independently, separately, or in decoupled phases.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the phrase "wherein the decoder neural network and the matching layer are obtained through independent training" characterizes the structural and functional operational state of the transceiver neural network parameters (i.e., that parameter updates or training phases of the decoder neural network and matching layer are executed in decoupled/separate optimization phases or under independent loss/freezing regimes), regardless of the physical location or specific execution hardware. Therefore, configuring Xie's receiver decoder network and transmitter channel encoder (matching layer) to be obtained through independent training phases as taught by Dorner directly satisfies this limitation.)
Xie and Dorner are analogous to the claimed invention as they are from the same field of endeavor of deep learning-based physical layer wireless communication systems and end-to-end autoencoders. Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to combine the Transformer-based semantic autoencoder transceiver architecture of Xie with the decoupled/independent training mechanism of Dorner.
The motivation is as recited by Dorner (
Dorner, Pg1, RC, P2, Lines 6-12, “Therefore, we show that training of GANs only requires a single dataset that can be collected in a one-shot transmission while this is not possible for RL-based training. Thus, training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter, while in RL training, a continuous feedback of the reward must be ensured”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
) such that executing independent/decoupled training between the receiver decoder neural network and the transmitter matching layer enables the system to adapt local parameters to changing physical channel environments without requiring full, continuous joint re-training of all transceiver components, thereby dramatically reducing training signaling overhead, computational latency, and operational complexity while maintaining optimal end-to-end decoding performance across dynamic wireless channels.
As to dependent Claim 15,
The combination of Xie and Dorner teaches all the limitations of Claim 14 as set forth above.
Xie does not explicitly teach
sending first indication information to the transmit end, wherein the first indication information indicates the transmit end to update a parameter of the matching layer
In the same field of endeavor, Dorner further teaches the apparatus according to claim 14, wherein the operations further comprise:
sending first indication information to the transmit end, wherein the first indication information indicates the transmit end to update a parameter of the matching layer (
Dorner, Pg1, RC, P3, Lines6-9, "...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter..."
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses a wireless communication autoencoder system wherein channel evaluation/training is performed at the receive end (or local receiver instance) and feedback information comprising updated transmitter weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) is sent back from the receive end to the transmit end to instruct, prompt, or cause the transmit end to update its neural network parameters (matching layer / transmitter parameters) to adapt to the physical channel environment.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the phrase "first indication information" is reasonably interpreted to encompass any feedback signal, control message, or updated parameter data (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) sent from a receive end that informs, instructs, or causes the transmit end to update or refresh its neural network parameters. Furthermore, applying Dorner’s receiver-to-transmitter weight update feedback mechanism to the split-autoencoder architecture of Xie (which maintains a frozen semantic encoder
PNG
media_image15.png
25
29
media_image15.png
Greyscale
and an adaptable channel encoder
PNG
media_image17.png
22
28
media_image17.png
Greyscale
) directly results in sending indication information to the transmit end to update the parameters of its channel encoder (which corresponds to the claimed matching layer under BRI).
Therefore, it would have been obvious to a person having ordinary skill in the art (POSITA), before the effective filing date, to configure the receive end apparatus of Claim 14 to send first indication information (e.g., updated transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
or feedback control signaling) to the transmit end as taught by Dorner, in order to enable the receive end to dictate or trigger transmitter matching layer parameter updates without requiring continuous real-time closed-loop reward feedback, thereby dramatically reducing feedback signaling overhead, transmission latency, and operational complexity while maintaining optimal end-to-end decoding performance across dynamic channels.)
As to dependent Claim 16,
The combination of Xie and Dorner teaches all the limitations of Claim 15 as set forth above.
Xie does not explicitly teach
sending the first indication information to the transmit end when a parameter representing a channel change degree is less than or equal to a first threshold
In the same field of endeavor, Dorner further teaches the apparatus according to claim 15, wherein the operations further comprise:
sending the first indication information to the transmit end when a parameter representing a channel change degree is evaluated or determined (
Dorner, Pg3, LC, Lines16-22, “To estimate the SNR(signal-to-noise-ratio) required by the demapper, we first calculate the error vector mangnitude (EVM) between the originally sent symbols and the equalized received symbols. We then calculate an average SNR per sub-carrier, defined by the mean over the EVM, and feed this SNR estimation to the demapper of the corresponding sub-carrier.”
Pg4, RC, Section IV, Lines10-13, “...we needed to add a pre-processing step that zero-forces the phase
PNG
media_image20.png
19
20
media_image20.png
Greyscale
of the first sub-carrier in time domain, as due to slight carrier frequency offset (CFO) we noticed random
PNG
media_image20.png
19
20
media_image20.png
Greyscale
for different measurements.”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses measuring, calculating, and evaluating channel change parameters or degradation metrics (such as EVM, CFO phase offset
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, or subcarrier SNR) at the receive end to determine parameter adaptation, and sending feedback/indication information comprising updated transmitter parameters (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) to the transmit end when such channel changes occur).
However, Dorner does not explicitly teach (Highlighted part)
sending the first indication information to the transmit end when a parameter representing a channel change degree is less than or equal to a first threshold.
In the same field of endeavor of wireless communication adaptive control systems, evaluating a measured channel change parameter (such as EVM, CFO phase offset
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, or subcarrier SNR degradation) against a predefined decision threshold to trigger sending first indication information (e.g., updated transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
or parameter update signaling) when the channel change degree is less than or equal to a first threshold represents a conventional design choice, routine optimization (MPEP 2144.05), and standard engineering practice well known to a person having ordinary skill in the art (POSITA).
Specifically, prior to the effective filing date, a POSITA aiming to optimize feedback signaling efficiency in Dorner’s transceiver system would have naturally recognized that continuous feedback signaling for minor channel fluctuations introduces unnecessary signaling overhead, processing delays, and network congestion. Applying a conventional decision threshold boundary, where sending the first indication information to trigger localized matching layer parameter updates occurs specifically when the measured channel change degree remains within a manageable bound (less than or equal to a first threshold), represents a routine design choice in closed-loop communication control to balance feedback signaling overhead against link alignment stability.
Therefore, it would have been obvious to a POSITA before the effective filing date to configure the receive end apparatus of Claim 15 to send the first indication information to the transmit end when a parameter representing a channel change degree is less than or equal to a first threshold as taught by Dorner in view of routine wireless communication control practices.
As to dependent Claim 17,
The combination of Xie and Dorner teaches all the limitations of Claim 14 as set forth above.
Xie does not explicitly teach
sending second indication information to the transmit end when a parameter representing a channel change degree is greater than a first threshold, wherein the second indication information indicates the transmit end to retrain the encoder neural network.
In the same field of endeavor, Dorner further teaches about the apparatus according to claim 14, wherein the operations further comprise:
sending indication information to the transmit end to trigger or effectuate retraining of the encoder neural network (
Dorner, Pg1, RC, P2, Lines9-11, "...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter...
Pg1, RC, P2, Lines12-14, "...This can be further combined with online retraining through label-recovery via outer channel codes at the receiver..."
Pg2, RC, P1, Lines4-7, "...after deployment on an actual channel, the channel gradient is unknown and only
PNG
media_image27.png
21
24
media_image27.png
Greyscale
can be updated straightforwardly using backpropagation, described as finetuning..."
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses executing retraining of encoder neural networks / transmitter weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) over a learned channel model and sending feedback/indication signaling from the receive end to the transmit end comprising updated transmitter parameters (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) to trigger, indicate, or effectuate retraining and parameter updating of the encoder neural network.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the phrase "second indication information" is reasonably interpreted to encompass any feedback message, control signal, or updated parameter payload sent to the transmit end that prompts, instructs, or triggers retraining of the encoder neural network.)
However, Dorner does not explicitly teach (Highlighted Part)
sending second indication information to the transmit end when a parameter representing a channel change degree is greater than a first threshold
In the same field of endeavor of wireless communication adaptive control systems, evaluating a measured channel change parameter (such as EVM, CFO phase offset
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, or SNR degradation) against a predefined decision threshold to trigger sending second indication information (e.g., retraining request or encoder update signaling) specifically when the channel change degree exceeds a first threshold (greater than a first threshold) represents a conventional design choice, routine optimization (MPEP 2144.05), and standard engineering practice well known to a person having ordinary skill in the art (POSITA).
Specifically, prior to the effective filing date, a POSITA aiming to optimize computational load and signaling efficiency in Dorner’s transceiver system would have naturally recognized that full retraining of encoder neural networks is computationally intensive and introduces significant processing latency. Applying a conventional decision threshold boundary, where full encoder retraining (triggered by sending second indication information) is reserved specifically for instances where the measured channel change degree exceeds manageable bounds (greater than a first threshold), represents a routine design choice in closed-loop communication control to balance dynamic link re-convergence against processing overhead.
Therefore, it would have been obvious to a POSITA before the effective filing date to configure the receive end apparatus of Claim 14 to send second indication information to the transmit end when a parameter representing a channel change degree is greater than a first threshold as taught by Dorner in view of routine wireless communication control practices.
As to dependent Claim 18,
The combination of Xie and Dorner teaches all the limitations of Claim 15 as set forth above.
Xie does not explicitly teach:
receiving third indication information from the transmit end, wherein the third indication information indicates the receive end to retrain the encoder neural network.
In the same field of endeavor, Dorner further teaches about the apparatus according to claim 15, wherein the operations further comprise:
exchanging training control signaling between the receive end and the transmit end to trigger or execute online/offline retraining of the encoder neural network (
Dorner, Pg1, RC, P2, Lines 9-14, "...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter... This can be further combined with online retraining through label-recovery via outer channel codes at the receiver..."
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
Pg5, RC, Section V, Lines9-11, "...non-stationary channels, i.e., the inability to converge for highly random or dynamic channels, which we consider as key subject of possible future work."
wherein Dorner explicitly discloses executing retraining of encoder neural networks/transmitter weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) when localized parameter updates are insufficient or fail to achieve link convergence across dynamic/non-stationary channel conditions, and exchanging control/status signaling between transceivers to coordinate retraining of the encoder neural network.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the phrase "third indication information" is reasonably interpreted to encompass any feedback message, control signal, status flag, or notification payload received from a transmit end that informs, instructs, or prompts the receive end that localized parameter adaptation has failed to achieve loss convergence and that retraining of the encoder neural network is required.)
However, Dorner does not explicitly recite the literal phrase "receiving third indication information from the transmit end".
It would have been obvious to a person having ordinary skill in the art (POSITA) before the effective filing date to configure the receive end apparatus of Claim 15 to receive status notification signaling (third indication information) from the transmit end indicating that localized matching layer parameter updates have failed to converge and prompting the receive end to execute retraining of the encoder neural network as taught by Dorner.
The technical motivation for incorporating this status notification feature is directly recited and established by Dorner (
Dorner, Pg5, RC, Section V, Lines9-11, "...non-stationary channels, i.e., the inability to converge for highly random or dynamic channels, which we consider as key subject of possible future work."
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”)
Specifically, Dorner explicitly identifies that localized parameter adaptation suffers from an operational limitation, namely, the "inability to converge for highly random or dynamic channels." A POSITA aiming to maintain closed-loop state synchronization in Dorner’s transceiver system would have directly recognized that when localized matching layer adaptation at the transmit end fails to achieve loss convergence under non-stationary channel fluctuations, continuing ineffective localized iterations causes state mismatch between the transmitter encoder and receiver decoder, resulting in frame decoding failure.
To overcome this explicit convergence limitation taught in Dorner, transmitting notification signaling (third indication information) from the transmit end to inform the receive end of non-convergence and trigger fallback to full encoder neural network retraining is technically necessary to halt invalid localized transmission, prevent transceiver state divergence, and promptly re-establish dynamic link convergence. Therefore, configuring the receive end apparatus of Claim 15 to receive third indication information from the transmit end indicating the receive end to retrain the encoder neural network would have been obvious to a POSITA before the effective filing date in view of Dorner’s explicit teaching of convergence limitations and closed-loop fallback retraining practices.
As to dependent Claim 19,
The combination of Xie and Dorner teaches all the limitations of Claim 15 as set forth above.
Xie does not explicitly teach:
receiving fourth indication information from the transmit end, wherein the fourth indication information indicates that training of the encoder neural network is completed.
In the same field of endeavor, Dorner further teaches about the apparatus according to claim 15, wherein the operations further comprise:
exchanging signaling between the receive end and transmit end indicating the completion and deployment status of retraining the encoder neural network (
Dorner, Pg1, RC, P2, Lines 9-11, “...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter...”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses completing retraining of transmitter weights / encoder neural networks (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) offline over a learned channel model and updating/deploying the final updated transmitter weights (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) to resume operational transmission across the channel.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification, the phrase "fourth indication information" is reasonably interpreted to encompass any completion flag, deployment signal, state acknowledgment message, or updated parameter weight payload (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) received from a transmit end that informs or indicates to the receive end that full retraining of the encoder neural network has concluded and that the transceiver pair can resume synchronized data decoding.)
However, Dorner does not explicitly recite the literal phrase "receiving fourth indication information from the transmit end".
It would have been obvious to a person having ordinary skill in the art (POSITA) before the effective filing date to configure the receive end apparatus of Claim 15 to receive signaling (fourth indication information) from the transmit end indicating that retraining of the encoder neural network is completed as taught by Dorner.
Specifically, a POSITA aiming to ensure operational resynchronization in Dorner’s transceiver system would have naturally recognized that following the execution of an offline or fallback retraining procedure for the encoder neural network, the receive end must be notified of retraining completion before active payload transmission can safely resume. Receiving a completion indication or parameter deployment payload (fourth indication information) from the transmit end upon conclusion of neural network retraining represents a routine state-transition acknowledgment in closed-loop communication protocol design to align the operational state of the decoder network with the newly updated encoder neural network, thereby preventing frame decoding errors and loss of frame synchronization.
Therefore, configuring the receive end apparatus of Claim 15 to receive fourth indication information from the transmit end indicating that training of the encoder neural network is completed would have been obvious to a POSITA before the effective filing date in view of Dorner’s explicit teaching of weight deployment and transceiver state resynchronization practices.
As to dependent Claim 20,
The combination of Xie and Dorner teaches all the limitations of Claim 15 as set forth above.
Xie does not explicitly teach:
sending first information to the transmit end, wherein the first information comprises information for updating the parameter of the matching layer.
In the same field of endeavor, Dorner further teaches about the apparatus according to claim 15, wherein the operations further comprise:
sending feedback data (first information) from the receive end to the transmit end comprising channel metrics or updated weight parameters used to update the transmit-end parameters (
Dorner, Pg1, RC, P2, Lines 9-11, “...training can be fully done at the receiver and only the updated transmitter weights have to be sent back to the transmitter...”
Dorner, Pg3, LC, Lines16-22, “To estimate the SNR(signal-to-noise-ratio) required by the demapper, we first calculate the error vector mangnitude (EVM) between the originally sent symbols and the equalized received symbols. We then calculate an average SNR per sub-carrier, defined by the mean over the EVM, and feed this SNR estimation to the demapper of the corresponding sub-carrier.”
Pg4, RC, Section IV, Lines10-13, “...we needed to add a pre-processing step that zero-forces the phase
PNG
media_image20.png
19
20
media_image20.png
Greyscale
of the first sub-carrier in time domain, as due to slight carrier frequency offset (CFO) we noticed random
PNG
media_image20.png
19
20
media_image20.png
Greyscale
for different measurements.”
Pg5, LC, Last 6 lines, “...while the WGAN-based setup was updated offline over the learned WGAN channel and the final transmitter weights
PNG
media_image18.png
29
30
media_image18.png
Greyscale
were updated only once. In terms of deployment complexity, the WGAN-setup thereby dramatically reduces transmission and weight update overhead of end-to-end training over the actual channel.”
wherein Dorner explicitly discloses obtaining feedback data at the receive end—such as CFO phase offsets
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, per-subcarrier EVM/SNR metrics, or updated transmitter parameters (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
)—and sending this feedback information (first information) from the receive end back to the transmit end to enable, instruct, or effectuate updating the parameter of the matching layer/transmitter neural network.
Under the Broadest Reasonable Interpretation (BRI) in light of the specification (specifically paragraphs [0135]-[0137] and Claim 13), the phrase "first information" is reasonably interpreted to encompass any feedback payload, channel quality metrics, CFO phase offsets, EVM/SNR estimates, received feature offset vectors, or updated parameter weight payloads (
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) transmitted from a receive end to a transmit end that provides data utilized by the transmit end to update or refresh its matching layer parameters.)
However, Dorner does not explicitly recite the literal phrase "sending first information to the transmit end".
It would have been obvious to a person having ordinary skill in the art (POSITA) before the effective filing date to configure the receive end apparatus of Claim 15 to send first information (e.g., CFO phase offsets
PNG
media_image20.png
19
20
media_image20.png
Greyscale
, EVM metrics, or updated transmitter weight payloads
PNG
media_image18.png
29
30
media_image18.png
Greyscale
) to the transmit end as taught by Dorner.
Specifically, a POSITA aiming to facilitate localized parameter adaptation in Dorner’s transceiver system would have naturally recognized that in order for the transmit end to update its matching layer parameters to match a dynamic channel environment, the receive end must feed back the requisite channel condition data or updated parameter payloads (first information). Transmitting channel measurement metrics or parameter weight payloads from the receive end to the transmit end provides the necessary informational input for the transmit end to execute localized parameter updates, thereby enabling link adaptation without requiring continuous, computationally intensive full-system joint retraining.
Therefore, configuring the receive end apparatus of Claim 15 to send first information to the transmit end for updating the parameter of the matching layer would have been obvious to a POSITA before the effective filing date in view of Dorner’s explicit teaching of receiver-to-transmitter feedback mechanisms and localized weight adaptation practices.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to DONG YOON JUNG whose telephone number is (571)270-0198. The examiner can normally be reached 8am-5pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Cesar Paula can be reached at (571) 272-4128. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/DONG YOON JUNG/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145
Prosecution Insights